Serum- and Glucocorticoid-induced Kinase Sgk1 Directly Promotes the Differentiation of Colorectal Cancer Cells and Restrains Metastasis

This article is featured in Highlights of This Issue, p. 451

The unit structure of the normal large bowel is the crypt. Stem cells at the crypt base produce progeny that expand in number and differentiate as they pass to the lumen into which they are shed. Although the biology of the stem cell niche is relatively well understood, much less is known about the process of differentiation. The involvement of signaling pathways such as Wnt, TGFβ/BMP/Hedgehog, Notch, and Ephrins is required for the production of main differentiated cell types (colonocyte, neuroendocrine, goblet, Paneth). However, especially for colonocytes, it is unclear to what extent differentiation is a fundamentally passive process (e.g., resulting from absence of microenvironmental Wnt activity, which declines as cells move up the crypt) or an active process, in which unknown pathways are switched on to drive the expression of genes involved in the functioning of the differentiated cells.

The degree of differentiation (usually termed “grade”) of colorectal cancers is a well-established tumor feature that is associated with patient prognosis, and its assessment forms part of the American Joint Committee on Cancer (AJCC) recommendations (1). The 5-year cumulative survival for patients with well-differentiated colorectal cancers is 80.9%, compared with 76.9% and 45.5% for moderately and poorly differentiated lesions, respectively (2). Because patients with colorectal cancer rarely die from effects of the primary cancer, poor differentiation must ultimately increase the risk of developing metastatic disease.

The molecular pathways that determine colorectal cancer differentiation are incompletely described. There are two main theories to account for how an undifferentiated tumor might arise: (i) cancer cells are selected for their ability to downregulate differentiation gene(s), the so-called “dedifferentiation” hypothesis; or (ii) differentiation reflects the cell of origin of the cancer in the crypt (3, 4). However, while it is evident that colorectal cancers demonstrate histologic and growth characteristics similar to undifferentiated or stem cells in normal tissue, neither hypothesis fully explains why colorectal cancers vary in their degree of differentiation.

Serum/glucocorticoid-regulated kinase 1 (SGK1) is a highly conserved member of the AGC family of serine/threonine kinases. It was originally identified as being transcriptionally activated in response to glucocorticoid stimuli in a mammary epithelial cell line (5). SGK1 is also induced by various cell stresses and is involved in the cell survival response. Its transcription is influenced by different types of hormones, growth factors, cytokines, and conditions of cellular stress (6). SGK1 also regulates a number of other ion channels, including K⁺, Ca²⁺, and Cl⁻ channels, and glucose transporters such as GLUT1 and SGLT1, thus controlling cell volume and osmolality (6). In the colorectum, such proteins are expected to be important in the functioning of differentiated colonocytes, as these cells primarily have a water-absorbing role. In keeping with this notion, we have previously demonstrated that SGK1 expression in normal intestinal epithelium is highest at the top of the crypt, corresponding to the positions of differentiated colonocytes, with reduced expression in intestinal adenomas (7).

Much remains unknown about SGK1′s function, and it appears to have a number of tissue-specific effects in both normal cells and cancers. SGK1 has been proposed to form an alternative arm of PI3K signaling to the canonical AKT arm. AKT and SGK1 have very similar catalytic domains, recognizing similar substrate motifs in vitro, and are both activated in a PI3K-dependent manner. This means that SGK1 potentially provides a parallel pathway to PI3K/AKT signaling and an alternative player in tumorigenesis (8, 9). SGK1 effects PI3K signaling through mTORC2 (10). SGK1 is phosphorylated, and therefore activated, by PDKs. SGK1 expression is also regulated at the mRNA level, for example, by Hippo pathway members Yap and Taz (11), and by glucocorticoid and mineralocorticoid receptors. Downstream, SGK1 phosphorylates the E3 ubiquitin ligase NEDD4-2, thus tagging it for degradation and allowing sodium channel expression. Other SGK1 targets include Foxo3, NRDG1, and MNK1. Nuclear SGK1 may also play a role in the epigenetic control of gene expression (12). SGK1 activation has been proposed as a mechanism of cancer resistance to PI3K inhibitors (13). There have been few studies of SGK1 in colorectal cancer. The kinase inhibitors SI113 and EMD638683 have anti-SGK1 activity and have been demonstrated to inhibit the growth of RKO and CACO2 colorectal cancer cells, respectively (14), often in combination with genotoxic therapies (15). Sgk1 knockout has also been reported to reduce intestinal tumor numbers in mice with germline Apc mutations (16).

In this study, we examine the role of SGK1 in intestinal cell differentiation, both in the normal crypt and in colorectal cancers. We determine the consequences of SGK1 reexpression for colorectal cancer behavior in model systems in vitro and in vivo, and identify c-Myc inhibition and plakophilin 3 expression as important downstream mechanisms. Finally, we test for associations between SGK1 levels and the clinicopathologic features of a set of sporadic colorectal cancers. Our findings suggest that the level of SGK1 expression is a strong determinant of colorectal cancer differentiation status, supporting the cancer “dedifferentiation” hypothesis. We also demonstrate that the differentiation status of a cancer is strongly associated with its metastatic potential. Our findings potentially open new avenues for research into colorectal cancer metastasis.

Human colorectal cancer cell lines were obtained from collaborators, and are commercially available from the ATCC: HT29 (ATCC HTB-38), LS174T (ATCC CL-188), SW480 (ATCC CCL-228). Murine MC-38 cell line is commercially available from Kerafast (ENH204). All colorectal cancer cell lines were grown in DMEM. Media were supplemented with 10% FCS and 1% penicillin/streptomycin, and incubated at 37°C and 5% CO₂. For sodium butyrate (NaBut) treatment, cells at 80% confluence were treated with 5 mmol/L NaBut diluted in medium. Mycoplasma status was tested prior to each experiment using the Lonza Mycoalert Mycoplasma Detection Kit (LT07-218).

A 1.7-kb region upstream of the start codon of rat Sgk1 was cloned into the pGL3-enhancer vector. Cells were transiently transfected with either of the Sgk1 promoter–pGL3 constructs or the Renilla luciferase pGL4.75 vector (Promega), as a control for transfection efficiency. Twenty-four hours after transfection, experimental cells were treated with 5 mmol/L NaBut and luciferase activity was measured 48 hours later with the Dual-Glo Luciferase Assay System (Promega).

For lentivirus production, a 3X-FLAG tag was added to the 3′ end of the full coding sequence of human SGK1. Transcription of SGK1 sequences was initiated by the CMV promoter, regulated by the Tet operator sequence in a Tet-Off–regulated manner.

Cells were harvested in lysis buffer supplemented with protease and phosphatase inhibitors (Roche). Protein content was measured by BCA assay (Pierce) and equal amounts were run onto pre-cast 4%–12% NuPage gels (Invitrogen). Blotted membranes were blocked in 5% skimmed milk in PBS and incubated with the indicated primary and secondary antibodies.

Cells were grown in confocal dishes with cover glass bottom (PAA), fixed in 3.7% paraformaldehyde for 10 minutes at room temperature, and permeabilized with 0.1% TritonX-100 in PBS for 5 minutes at room temperature. Fixed cells were incubated either with serum obtained from the species in which the secondary was raised or with 1% BSA to prevent nonspecific binding. Cells were then incubated with primary antibodies for 1 hour, rinsed, incubated with fluorescently labeled secondary antibodies for 1 hour in the dark, and finally washed and counterstained with DAPI (Sigma). Samples were analyzed with the ZEISS 510 MetaHead confocal microscope.

Tissue microarrays (TMA; CO484a, CO485) were obtained from Insight Biotechnology Ltd, the UK suppliers of US Biomax. They comprised histologic sections of colon cancer specimens with information on TNM and clinical stage, differentiation, and pathologic grade. Scoring for SGK1 protein expression was performed following IHC on a scale of 1–5 for each specimen.

Endoscopic biopsies were taken from the descending colon of five normal control patients and were microdissected into crypt tops and bottoms. RNA was extracted, amplified, and quantified using Illumina Beadchip microarrays. Analysis was performed using the R package “limma.”

RNA was isolated from pelleted cells using the Qiagen Rneasy MinElute clean-up kit and analyzed using Tapestation. RNA-sequencing libraries were prepared using the QuantSeq 3′ mRNA library prep kits (Lexogen) and sequencing performed on an Illumina Hiseq 4000 with 75-bp paired-end reads. RNA sequencing for the colorectal cancer cell lines were performed in duplicate and for animal experiments in triplicate. Sequencing was performed to an average depth of 5 million reads. Reads were demultiplexed and mapped using Hisat2 to the reference human genome GRCh37. Feature counting was performed using Htseq version 0.6.1 and read normalization and differential expression, using DESeq version 1.10.1. JavaGSEA was used using the Molecular Signatures Database on preranked lists to perform gene-set enrichment analysis. Curated signatures for cellular senescence, apoptosis, and proliferation were obtained from published genesets (17).

Cell proliferation was measured with the CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega). Apoptosis was measured with the Caspase-Glo 3/7 Assay (Promega) under basal conditions and upon osmotic stress stimulus, achieved by treating the cells with 500 mmol/L sorbitol (Sigma) for 24 hours.

Anchorage-independent growth was measured as the ability of cells to form colonies in soft agar. Briefly, a bottom layer consisting of 0.6% agar (Sigma, A9414) diluted in tissue culture medium was plated out in 6-well plates. Equal numbers of cells were diluted in 0.3% agar in tissue culture medium and plated onto the bottom layer at a concentration of 1,000 cells per well, in triplicates. Cells were incubated for three weeks and fed with 200 μL of medium twice weekly. Images were acquired with a scanner and the number of colonies formed was counted for each well.

Transwell assays (8 μm) were obtained from Corning and 3 × 10⁵ cells were trypsinized and suspended in DMEM with 0% FBS and placed in the top chamber. DMEM with 10% FBS was placed in the bottom chamber. Sixteen hours later, cells were removed from the top chamber and cells on the bottom surface of the transwell assay were stained using Kwik-Diff stains (Thermo Fisher Scientific). Automated cell counting was performed using the Nikon NIS Elements software package.

Stable isotope labeling of amino acids in culture (SILAC) was performed with the SILAC Protein Quantitation Kit (Thermo Fisher Scientific/Pierce) according to instructions. HT29-SGK1 cells were cultured in “heavy” medium (13C6Lys + 13C6/15N4Arg) and HT29 parental cells in “light” medium in the presence of doxycycline for 5 passages, after which incorporation of the heavy label was measured by tandem mass spectrometry. Resulting peptides were analyzed by LC-MS/MS using an Ultimate3000 HPLC system coupled on-line to a 3D high-capacity ion trap tandem mass spectrometer via a pneumatically assisted nano-electrospray source as described previously (18).

Animals were a mix of 5- to 10-week-old male and female JAX NSG mice (NOD.Cg-Prkdcscid Il2rgtm1WjI/SzJ) obtained from Charles River Laboratories. Animals were maintained under specific pathogen-free conditions, and food and water were supplied ad libitum. Housing and all procedures involving animals were performed according to protocols approved by the United Kingdom Home Office, in compliance with Animals (Scientific Procedures) Act and home office guidelines on animal welfare. Orthotopic implantation of tumors was performed on 5- to 10-week-old mice under anesthesia by isoflurane inhalation, using a 30-gauge custom-made needle under direct endoscopy vision with the Coloview system (Karl Storz; ref. 19).

The human cohort data were obtained from The Cancer Genome Atlas (TCGA), which is publicly available for the scientific research community. Ethical approval for the data was obtained from the NIH (Bethesda, MD).

To confirm that expression of SGK1 alters along the crypt axis of the colon and small intestine, endoscopic biopsies from the terminal ileum and colon from 5 patients were obtained and microdissected into crypt tops and bottoms. RNA was extracted and global gene expression was quantified using microarrays. SGK1 expression was 9.6-fold higher in small-bowel villi than crypts (P = 0.003) and 10.8-fold higher in colonic crypt tops than bottoms (P = 0.1), consistent with previous data showing that SGK1 is expressed predominantly in differentiated enterocytes and colonocytes. Increased expression of SGK1 was associated with increasing expression of SHBG, CIDEC, DUSP5, and MAF, known drivers of differentiation in cell lines. DUSP5 was especially interesting, as low expression is known to be associated with poor tumor differentiation, lymph node metastasis, distant metastasis, and poorer overall survival (20). Expression of SGK1 was negatively associated with expression of ASCL2, OLFM4, EPHB2, RETNLB, MSI1, and SOX9, genes associated with stem cell maintenance and function (Supplementary Table S1).

Because colorectal cancers are typified by reduced differentiation, we compared the levels of SGK1 mRNA in colorectal cancers with those in normal intestine (data from The Cancer Genome Atlas). SGK1 expression was >10-fold lower in the former (Supplementary Fig. S1). In keeping with this, SGK1 was recently identified as a “hub gene”, the reduced expression of which is central to a set of gene expression changes in colorectal carcinogenesis (24).

NaBut is a short-chain fatty acid normally found in the colorectal lumen. NaBut promotes cellular differentiation in vitro in a number of colorectal cancer cell lines (22, 26). Because our data showed SGK1 to be upregulated in differentiated cells of the intestine, at the crypt tops, we investigated whether SGK1 expression was induced by NaBut-driven differentiation in colorectal cancer lines. Two cell lines were selected, LS174T, which expresses almost no endogenous SGK1, and HT29, which expresses modest levels (23). Upon NaBut treatment, both HT29 and LS174T cells showed a significant increase in SGK1 transcript levels over time, as measured by qRT-PCR (P < 0.01; Fig. 1A). This increase was confirmed in luciferase reporter assays (3.9-fold for HT29, 1.7-fold for LS174T; data not shown).

The NaBut experiments demonstrated that SGK1 is induced by differentiating agents. On the basis of these data and the SGK1 expression pattern in the crypt, we pursued the hypothesis that SGK1 is a key player in signaling pathways that induce and directly control differentiation of colorectal cells. To determine experimentally whether this could be the case, we developed a system to establish stable, inducible expression of SGK1 using a lentivirus-based system. HT29 and LS174T colorectal cancer cell lines were used as models and comparisons made to their control counterparts that had been transduced with an empty lentivirus (HT29-SGK1 vs. HT29-WT, LS174T-SGK1 vs. LS174T-WT). This system resulted in levels of SGK1 mRNA expression that were comparable with those in normal intestinal tissues (Supplementary Fig. S2) and in the production of SGK1 protein (Supplementary Fig. S3). Taking this approach, it was possible to demonstrate that functionally active SGK1 protein was expressed, based on reduced levels of the known SGK1 phosphorylation target, NEDD4-2 (Supplementary Fig. S4).

The process of differentiation in colonocytes is characterized by several histologic features, including the formation of junctional complexes that maintain cell shape and the establishment of basal-apical cell polarity, with the formation of actin filament caps that colocalize with apical markers in differentiated colonocytes (24). Upon NaBut-induced differentiation, LS174T cells are known to undergo specific morphologic changes, which include the appearance of mucin-like granules (21). We found that similar morphologic changes were induced by reexpression of SGK1 in the absence of NaBut, and there was an increased tendency for cells to grow as colonies. HT29 cells treated with NaBut or in which expression of SGK1 was induced, showed comparable morphologic changes (Supplementary Fig. S5).

We investigated whether other features of differentiation could be identified in our colorectal cancer cell lines following reexpression of SGK1, specifically investigating the levels and distribution of proteins important for the formation of key junctional proteins, β-catenin, ZO-1, and E-cadherin. SGK1 caused an increase in total ZO-1 but not β-catenin (Fig. 1B; Supplementary Fig. S6). Immunofluorescence analysis of E-cadherin demonstrated increased membrane expression following SGK1 reexpression, and this was confirmed following cell fractionation Western blotting (Fig. 1C). We then assessed markers of cell polarity, which is a feature of differentiated intestinal epithelial cells. We found that LS174T and HT29 displayed actin caps upon reexpression of SGK1, and these were not seen in control cells. Moreover, the apical marker villin, which showed diffuse cytoplasmic expression in control cells, colocalized with the actin caps (Fig. 1D). Together, these findings demonstrated the ability of SGK1 reexpression to induce colonocyte differentiation.

We also transiently expressed SGK1 in the HCT116 colorectal cancer cell line, which has very low endogenous SGK1 and does not spontaneously differentiate in culture, perhaps having irreversibly lost the ability to do so. Following transient high-level SGK1 expression, HCT116 cells ruptured, with no evidence of differentiation.

To understand the phenotypic changes in cells that are induced by SGK1, we performed a number of functional assays using the reexpression cell lines, HT29-SGK1 and LS174T-SGK1. Twenty-four hours after reexpression of SGK1, we observed a significant reduction of approximately 40% in apoptosis rates under basal conditions, and a more modest but significant reduction upon induction of osmotic stress with sorbitol (Fig. 2A). Apoptosis rates subsequently increased over the time course of the experiment to reach levels comparable with or exceeding those of control cells (data not shown), suggesting that the SGK1-induced protection was only temporary, or that resistant clones had grown out. SGK1 reexpression also led to modest, but significant, reductions in proliferation in the SGK1 reexpressing cells compared with controls (Fig. 2A), and this was maintained over time.

To probe the effects of SGK1 on stem-like cell properties, colonosphere formation assays were performed. Upon reexpression of SGK1, both HT29 and LS174T cell lines showed dramatically impaired formation of colonospheres (Fig. 2B), in keeping with SGK1-mediated reversion of the colorectal cancer cell lines to a more differentiated phenotype. To investigate the effects on anchorage-independent growth, soft-agar assays were performed to assess characteristics observed in differentiated cells, anoikis (cell death), and the inability to proliferate, when there is loss of adhesion from the extracellular matrix. SGK1 reexpression results in reduced colony formation (Fig. 2C).

Finally, we performed a Boyden migration assay that assesses cellular motility across a transwell membrane toward a growth factor gradient. At 24 hours, reexpression of SGK1 resulted in a significant reduction in rates of migration compared with cells that had been transduced with an empty lentiviral construct (Fig. 2D). This observation was confirmed using a different approach, where a migration assay was performed on 11 colorectal cancer cell lines that naturally expressed different amounts of SGK1. This assay identified that lines with low SGK1 mRNA expression, such as LS174T and SW480, exhibited the highest migration rate, whereas cell line with higher SGK1 (COLO205, SNU-C4, SW1222) demonstrated low cellular migration (Fig. 2E).

The Cancer Cell Line Encyclopedia (CCLE) was then utilized to obtain SGK1 expression for a total of 43 metastatic and primary colorectal cancer cell lines. Of the 42 cell lines, 9 were derived from metastatic sites. Primary colorectal cancer lines expressed significantly higher SGK1 than lines derived from metastases [2.7 transcripts per kilobase million (TPM) vs. 6.7 TPM; P < 0.05], but expression was both low and variable (Supplementary Table S2).

In summary, these findings point toward a link between increased SGK1 expression and reduced proliferation, apoptosis, stem cell phenotype, and motility.

SGK1 could in principle influence numerous gene networks through protein phosphorylation, thus amplifying environmental or other triggers for differentiation. To gain clues about the gene expression changes induced by SGK1, it was reexpressed in 7 colorectal cancer cell lines (GP2D, HT29, HCT116, LS174T, SW1222, SW480 and T84). Whole-transcriptome profiling was performed using 3′-RNA sequencing.

Three genes were found to be differentially expressed: SGK1, NEFM, and RNA18S5 (P_adj all <0.05). To determine whether there were more subtle changes in expression of gene sets, GSEA using signatures from the MsigDB databases and curated genesets were performed. Reexpression of SGK1 was associated with enrichment for gene networks involved not only in the “protein kinase cascade” and “stress-activated protein kinase signaling” as expected from existing data, but also for “cell polarity”, “differentiation,” and “cell matrix adhesion”. A focused analysis of cancer signaling pathways demonstrated that expression levels of genes in several pro-oncogenic pathways were significantly decreased following SGK1 reexpression. This effect was most pronounced for the gene sets “MYC targets” and “retinoblastoma/E2F signaling” (NES = −26.1 and −1.56 respectively, P < 0.001 for both; Fig. 3A; Table 1). Following reexpression of SGK1, expression of MYC mRNA was reduced by 23.6% (P = 0.03) and expression of c-Myc target PGK1 was reduced by 23.1% (P = 0.03).

In colorectal cancer, c-Myc lies at the crossroads of many growth-promoting signal transduction pathways and plays a key role as a prometastatic transcription factor, regulating epithelial-to-mesenchymal transition, cell–cell matrix interactions, and cellular invasion in cell and murine models. To confirm reduction in c-Myc protein, we transduced colorectal cancer lines HT29 and SW480 with SGK1 reexpression lentivirus, a kinase-dead SGK1 reexpression lentivirus or an empty lentiviral vector. Upon reexpression of SGK1, there was a consistent and significant decrease in c-Myc protein expression only in the cell lines carrying the SGK1 reexpression vector (mean reduction = 56% at day 4; Fig. 3B; Supplementary Fig. S7). Using antibodies specific to phospho-Ser62 and phospho-Thr58, we demonstrated that there was a corresponding decrease in phosphorylation at both c-Myc Ser62 and Thr58 in the SGK1 reexpressing cells (Fig. 3C).

We searched for proteins that were differentially expressed as a consequence of SGK1 reexpression and hence were potential downstream effectors of SGK1-induced differentiation. We utilized SILAC on our HT29-SGK1 and parental cells, followed by mass spectrometry differential analysis. The experiment identified a small number of proteins showing differential expression upon SGK1 reexpression. Mass spectrometry results consistently indicated an increased abundance (∼20-fold) of plakophilin 3 (PKP3; Fig. 4A), an essential component of desmosomes that has been linked to metastatic potential and differentiation in tumors (25, 26). Western blotting (Supplementary Fig. S8A) confirmed higher levels of PKP3 expression in both HT29-SGK1 and LS174T-SGK1 cells compared with parental cells. Investigation of PKP3 expression by IHC in normal human colon showed increased expression in differentiated cells, paralleling SGK1 expression (Fig. 4B).

To assess whether PKP3 mediates the effects of SGK1 on cell differentiation, we engineered lentiviral-based shRNA knockdown of PKP3 in HT29-SGK1 and LS174T-SGK1 cells. When compared with control cells reexpressing SGK1 and transduced with a nontargeting shRNA virus, abolition of PKP3 expression (Supplementary Fig. S8B) lowered membranous E-cadherin expression, and also restored colony-forming ability (Fig. 4C and D).

Our experiments demonstrated that SGK1 exerted diverse effects on the cancer cell phenotype. Many of the processes affected (differentiation, migration, anchorage-independent growth, intercellular adhesion, and c-Myc expression) are important in metastasis. We therefore developed an orthotopic xenograft mouse model of colorectal cancer metastasis, by endosocopically injecting genetically modified colorectal cancer cells into the submucosa of the mouse descending colon. We initially tested Sgk1 knockout in MC38 colorectal cancer cells, using a CRISPR/Cas9 strategy as part of a whole-genome screen in 8 mice. Sgk1 knockout did not detectably change metastasis rates, as might be expected given the very low Sgk1 levels in colorectal cancers (details not shown). We therefore tested the effects on metastasis of overexpressing SGK1. LS174T-SGK1 and LS174T-WT cells were labeled using a luciferase-expressing lentivirus (pHIV-Zsgreen-luc2) to enable in vivo imaging, and transplanted into Nod/Scid Gamma mice under endoscopic guidance. We found SGK1 mRNA expression in the LS174T-SGK1 cell lines and primary xenografts to be at levels comparable with normal epithelium (data not shown).

All animals (n = 16) developed tumors in the colonic lumen following successful implantation. By day 21, mice became symptomatic principally due to intestinal obstruction but also metastatic outgrowth. All mice developed lung metastases, but secondary spread to the liver was uncommon (4/16 mice) and lesions in bone and other distant sites were not detected. At day 21, mice were culled, lungs were resected and ex vivo analysis performed.

Ex vivo imaging demonstrated that there was a striking 25.8-fold decrease in lung metastasis burden (P < 0.008) in the mice that received LS174T-SGK1 compared with those that received LS174T parental cells. In a replication cohort (n = 15), there was a similar (20.4-fold) decrease in lung metastasis burden (P < 0.001; Fig. 5A and B). In contrast, there was no significant difference in primary tumor size between the LS174T-SGK1 and LS174T-WT groups (Supplementary Fig. S9A).

These findings are in keeping with SGK1-mediated differentiation in vitro not limiting primary tumor growth, broadly consistent with the modest decreases in proliferation found in vitro (see above), but acting potently as a suppressor of metastasis.

To determine the stage at which SGK1 exerted its effect on metastasis, whether inhibiting the ability of cells to migrate away from the primary tumor, or inhibiting later stages in metastasis such as the ability of cells to survive in the circulation and enter metastatic tissues, a tail-vein assay was performed. LS174T-SGK1 and LS174T-WT cells were injected into the tail veins of 8 Nod/Scid Gamma mice. Bioluminescence ex vivo analysis confirmed a marked reduction in metastasis in the LS174T-SGK1 cells, comparable in magnitude to the reduction seen in the orthotopic xenografts (P < 0.001; Supplementary Fig. S9B).

Having established that SGK1 expression could induce features of differentiation in colorectal cancer cell lines, we examined this association in human colorectal cancers. A tissue microarray of sporadic colorectal cancer resection specimens (n = 86) was obtained comprising of cancers of principally stage II/III colorectal cancer and from patients with a mean age of 51.7 years (Supplementary Table S3). Differentiation grading had been performed by accredited histopathologists according to AJCC guidelines. We examined SGK1 expression using IHC (Supplementary Fig. S10). This demonstrated that there was a decrease in SGK1 expression in poorly differentiated than well or moderately differentiated cancers (t test, P < 0.01; Fig. 5C). This result validated our experimental findings suggesting an association between low SGK1 expression and poor colorectal cancer differentiation in both colorectal cancer lines and tumor resection specimens.

Poor differentiation (high grade) is a well-known poor prognostic feature of colorectal cancer. Undifferentiated colorectal cancers are associated with increased cellular motility and features such as tumor and epithelial–mesenchymal transition (EMT; ref. 27). Patients with poorly differentiated colorectal cancers tend to be younger, have larger, right-sided tumors with lymphovascular invasion, a higher frequency of MSI, a significantly poorer 4 year overall survival (28) and a higher propensity to develop metastasis (29). As we have demonstrated the expression of SGK1 to be closely associated with colorectal cancer differentiation/grade, we sought to determine whether SGK1 expression could be utilized as a similar marker of poor prognosis.

The Cancer Genome Atlas sample set comprised an annotated set of U.S.-based patient genomes with colorectal cancer of various stages (30). A set of cohort of 284 patients had detailed follow-up information and RNAseq had been performed on all specimens, enabling an analysis of the effect of SGK1 expression. SGK1 expression was expressed as quintiles as it was non-normally distributed and quintiles used as a quantitative variable. This analysis found a significant association between lower SGK1 expression and poorer survival (P = 0.023; Fig. 5D). There was no correlation between SGK1 expression and tumor stage and the association of low expression of SGK1 with poorer prognosis was observed across all stages of colorectal cancer except the small number of patients with stage I disease (Supplementary Fig. S11).

SGK1 is a serine/threonine kinase that shares homology with proteins of the AKT family. To date, the role of SGK1 has been most clearly elucidated in regulation of ion channels (31), neuronal function (32), and T helper cell differentiation (33). More recently, there has been a growing realization that SGK1 lies at the junction of, and could mediate the effects of, several oncogenic signaling networks. The effects of SGK1 are variable, depending on the tissues or cell lines analyzed and the model systems used. A number of studies have highlighted the proproliferative effects of SGK1: in hepatocyte regeneration, SGK1 mediated the phosphorylation of ERK2 and the activation of the Ras-MAPK kinase pathway (34); and in breast cancer, activation of SGK1 sustained mTOR activation in the context of AKT inhibition through direct phosphorylation of TSC2 (35). SGK1 is also thought to act in conjunction with PHLPP1 to enable Ras-mutant cancers survive anoikis during metastasis (36). In contrast, in HEK293 cells, SGK1 inhibited the Ras-MAPK kinase pathway through phosphorylation of B-Raf, which is likely to exert an antiproliferative effect (37).

In terms of regulation of SGK1 expression, in colorectal cancer, mutations in SGK1 may be observed in up to 2% of cases, although the mutations are usually missense and not isolated to the kinase domain (38). Our previous work has identified that downregulation in colorectal cancers appears to be independent of promoter hypermethylation (39) and these observations suggest that control is likely to be transcriptionally regulated.

Our focus in this work was on the potential role of SGK1 in actively promoting differentiation, based on our validated finding that in the colorectum, SGK1 is expressed predominantly in the terminally differentiated epithelial cell compartment, in which cells have little proliferative potential, but are required to survive and function in a potentially hostile osmolar environment. We initially showed that SGK1 is induced by the extrinsic differentiating agent NaBut, consistent with NaBut acting, at least in part, through SGK1. On the basis of this preliminary evidence, we directly pursued the hypothesis that SGK1 reexpression induces features of differentiation. For this, we used both cellular and molecular read-outs in colorectal cancer cell lines that retain some differentiation capability. We found that SGK1 induces not only a gene expression profile of differentiation, but also several specific features of differentiation, including cell polarity, increased tight junction protein expression, and relocation of adherens junction proteins to the membrane. Stem cell–like features, proliferation, apoptosis, growth in soft agar, and migration were correspondingly decreased. Our data showed that mechanisms of SGK1 action plausibly include reduction of c-Myc signaling, and increased expression of the desmosome protein plakophilin 3. In a colorectal cancer cell line orthotopic xenograft model, we found SGK1 reexpression to reduce the growth of lung metastases. The results of this assay suggested that SGK1-mediated differentiation of colorectal cancer lines inhibits the later stages of metastasis, perhaps functioning by limiting the ability of cells to migrate into the lung parenchyma from the endothelial cells.

Some caveats must be applied to our data. First, SGK1 expression can be hard to assess using Western blots/IHC, at least in part because the protein has multiple isoforms and there are few antibodies with good specificities. We therefore relied on mRNA expression using a pan-isoform probe or total isoform expression by sequencing as a read-out for many of our assays. We did, however, perform Western blot analysis of tagged SGK1, a limited analysis of native SGK1, and also showed that functional SGK1 protein was produced on the basis of assessment of NEDD4-2 levels. In combination, these experiments strongly supported the existence of active SGK1 in our experiments. Second, given the low SGK1 levels in colorectal cancer cell lines (including those with differentiation capacity) and the potential redundancy between SGK1 and related SGKs, our experiments necessarily relied on SGK1 reexpression rather than knockdown. While we appreciate that overexpression of kinases can lead to nonphysiologic effects, SGK1 reexpression in our systems led to near-physiologic levels of mRNA. Finally, we have utilized an endoscopic transplantation model to determine effect on metastasis. The limitation of this approach is that as the lumen of the descending colon is narrow and mice become symptomatic relatively early. In contrast, the alternative, a cecal transplantation model allows mice to survive for a relatively longer period and often leads to increased levels of liver metastasis. We utilized the endoscopic transplantation model as it does not involve the creation of a track made from the external surface of the bowel and believe that, anatomically, the cecum of the mouse is markedly different from human.

One largely unanswered question in the field is the differences and similarities between the AKT and SGK1 arms of the PI3K pathway. We speculate that while both arms promote cell survival, which may explain some of the data showing that loss of SGK1 is deleterious to cells (see above), SGK1 does this in the context of its additional function of promoting differentiation, particularly in the intestine. In human colorectal cancers, features of differentiation are not necessarily lost, with well and moderately differentiated cancers retaining some ability to form glandular units. Tumor differentiation is an important histologic feature that is associated with poor cancer prognosis. The mechanisms through which the differentiation status of a tumor occurs and affects its behavior and patient prognosis have, however, been somewhat overlooked. Our findings are consistent with the reduced expression of SGK1 being one important means by which tumors may “dedifferentiate” and acquire other malignant phenotypes such as migration and invasion. We conclude that a renewed focus on therapeutic agents to differentiate tumors, rather than simply to target cell survival, is justified.

No potential conflicts of interest were disclosed.

Conception and design: L.Y.W. Lee, I. Tomlinson

Development of methodology: L.Y.W. Lee, C. Woolley, I. Tomlinson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.Y.W. Lee, C. Woolley, T. Starkey, S. Biswas, T.S. Mirshahi, S. Segditsas

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.Y.W. Lee, T. Starkey, S. Biswas, I. Tomlinson

Writing, review, and/or revision of the manuscript: L.Y.W. Lee, C. Woolley, T. Starkey, S. Biswas, I. Tomlinson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Biswas

Study supervision: C. Bardella, S. Irshad, I. Tomlinson

We would like to thank members of the Tomlinson lab for their help and advice. L.Y.W. Lee is supported by a Guts UK charity trainee award, a Doctoral training program (DTP) award, and the Medical Research Council (MRC). I. Tomlinson is supported by a Cancer Research UK (CR-UK) Programme Grant.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.